Quantifying the quality of localisation

 

Localisation is one of the fundamental tasks of any robotic system. Only reliable pose estimation allows a robot to execute its mission successfully. Therefore, for long-lasting autonomous missions, it is essential to equip the robot with the ability to estimate the trust in the current pose estimate.  The level of trust combined with pose error estimation can help to assure a long and safe operation of a mobile robot.

 

In recent years a substantial amount of work was given to improve the quality of MCL-based localisation methods and its robustness. However, there are no methods that will allow the robot to monitor and improve its localisation actively using one of the following methods:

- retreating to the last correct position

- switch between localisation method

- modify the localisation parameters

 

In the run of this thesis the student will execute the following tasks:

- Survey the state of the art

- Implement adaptable localisation system

- Implement tools for localisation error estimation

- Implement tools for localisation certainty estimation

- Run evaluation experiments

 

It is expected that the student has a good knowledge of localisation methods and probabilistic methods in robotics (Bayesian statistics). It is also expected that the student can work with hardware sensors (2D/3D LIDARs, RGB-D cameras), has excellent programming skills (C/C++) and experience in working with ROS.